DNA tests have made it clear that innocent people have been sent to prison after a witness picked them out of a lineup. In fact, since 1989, more than 70 percent of 333 wrongful convictions in the U.S. have been influenced by misidentification from eyewitnesses. But researchers recently reported that the disdain for eyewitness identification is not always warranted. They found that if witnesses shown a lineup for the first time are asked to state their confidence in their choice, the identifications they are most confident of are much more likely to be of the suspect than of the innocent.

"Ignoring low confidence in the beginning is a grave error,” says lead researcher John Wixted at University of California, San Diego. “The witness is telling you that there's a good chance they're making a mistake."

At the same time, the study, published in this week’s Proceedings of the National Academy of Sciences, investigates a longstanding debate on how to perform a lineup—which could effectively affect confidence levels in witnesses. When a police officer wants to see if a witness will pick out a suspect from a crowd, he or she will assemble a set of photos of people who match the description of the perpetrator. If the witness chooses a suspect, that's further evidence that the police are on the right track. The fear, though, is that witnesses will pick innocent people. Numerous studies through the past few decades have examined how to structure a lineup so as to minimize that possibility, and have settled on showing people the photos one by one, instead of all together. About 30 percent of the police departments in the US have adopted this sequential method, rather than the older simultaneous method.

Wixted says, though, that some studies used to make that decision overlook an important detail. They look at the ratio of suspect identifications to mistaken identifications of innocents, but assessing the success ratio alone, Wixted explains, without accounting for confidence doesn’t tell the whole story. Sometimes, a witness will just pick someone randomly or will openly state that they are not sure about their identification. When studies don't weight those guesses differently than confident statements, they aren't reflecting the way lineups are usually used, Wixted says. “In the real world, they often don't even count random guesses,” he says. “Some jurisdictions do, but [in] most places if the witness is hesitant, they won't take it to court.”

In the current study, the researchers examined lineups administered by the Robbery Division of the Houston Police Department to see how the simultaneous and sequential processes compared when witness were asked to rate their own confidence. The lineups, of which 187 were simultaneous and 161 sequential, were administered by people who themselves were unaware of the suspect’s identity, and only cases in which the suspect was a stranger to the witness were included. Witnesses rated their identification confidence as low, medium, or high.

In a third of the cases, the witnesses did not identify anyone. In another third, they identified the suspect, and in the remaining cases they chose someone who was not suspected, or a “filler.” When the researchers compared the confidence rates between the suspect identifications and the filler identifications, however, they found something interesting: Very few people who chose fillers were confident of their choice—most low confidence IDs, in fact, were of fillers. By contrast, most high-confidence identifications were of the suspect. That suggests that confidence is a good indicator of whether the person identified is the suspect.

Comparing the results of the two different lineup techniques, the researchers found that employing the simultaneous method produced more confident identifications, leading to the conclusion that simultaneous identification may actually be more useful to police departments than sequential identification.

The difference between the two methods is statistically very slight, however, notes Gary Wells, a professor of psychology at Iowa State University who studies eyewitness memory. “The more important part of this article is that witness confidence did a good job of helping sort between accurate and mistaken witnesses (regardless of whether it is simultaneous or sequential),” he wrote in an email. It does not matter so much which procedure a police department uses—what matters more is that they ensure the lineup is administered by someone who does not know who the suspect is and thus cannot influence the witness one way or another, and that they take the measure of the witness's confidence on the spot. “Police departments, jurors, judges need to know that if their jurisdiction is not using double-blind lineup procedures” in which the test administrator and witness have not been told which is the suspect, “then these findings do not apply to them,” he continues. Fewer than half of US police departments use a double-blind procedure, he writes.

The study is part of a body of research suggesting that witness confidence is not as unreliable as had sometimes been thought, provided the procedure is blind and the measure is taken right away. Studies have found that in a large proportion of cases where a witness confidently identified an innocent person as the culprit in court, the witness was not so sure at the initial lineup. 

The shift in favor of measuring confidence during a lineup has not yet made it beyond the realm of research. Wixted hopes his findings will influence how line-ups are handled by the police, but, he says, “the word, has not gotten out."